Correlation is Not Causation – Really Big Data Edition

Via Cosma in comments at the other place I hang out, this is a very nice teaching tool for the sole and single purpose of getting this point across to students. NB that you need to be logged into a Google ID to use it. My favorite so far is the .8222 correlation between my random graph and searches for “frogsex” (the mind squirbles).

2 Responses to Correlation is Not Causation – Really Big Data Edition

  1. Andreas Moser December 12, 2011 at 12:35 pm #

    I use this example to try to get people to memorize that correlation is not causation: http://andreasmoser.wordpress.com/2011/04/05/if-you-dont-read-my-blog-you-will-die/ – to no avail, I am afraid.

    • matt w December 13, 2011 at 11:43 am #

      I’m not even sure you’ve got correlation there. 150,000 people dying yesterday sounds like a lot, but it’s really a very small proportion. Unless you have 40,000 readers a day — which, with all due respect, I doubt — then odds are that none of your readers will die any given day, just by random chance. Have I got that right?